Autonomous robust assembly planning

Abstract

A method for tuning the force control parameters for a general robotic assembly operation. The method uses numerical optimization to evaluate different combinations of the parameters for a robot force controller in a simulation environment that is built based on a real-world robotic setup. This method performs autonomous tuning for assembly tasks based on closed loop force control simulation, where random samples from a distribution of force control parameter values are evaluated, and the optimization routine iteratively redefines the parameter distribution to find optimal values of the parameters. Each simulated assembly is evaluated using multiple simulations including random part positioning uncertainties. The performance of each simulated assembly is evaluated by the average of the simulation results, thus ensuring that the selected control parameters will perform well in most possible conditions. Once the parameters have been optimized, they are applied to real robots to perform the actual assembly operation.

What is claimed is: 
     
       1. A method for autonomously tuning controller parameters for a robotic assembly operation in which a robot manipulates a first part into an assembled position with a second part, said method comprising:
 initializing a distribution of the controller parameters to be tuned; 
 providing a plurality of random samples of the controller parameters from the distribution; 
 running a set of simulations of the assembly operation for each of the samples in the plurality, on a computer having a processor and memory, the simulations including a compliance controller model of a robot performing the assembly operation, where each set of simulations includes a plurality of simulations each using a different fixed pose deviation of the second part; 
 computing a cost function value for each of the simulations, and an average cost function value for each of the sets of simulations; and 
 running a numerical optimization to optimize values of the controller parameters, including; 
 redefining the distribution of controller parameters based on a quantity of the sets of simulations having the lowest average cost function values; 
 returning to providing the first plurality of random samples of the controller parameters from the distribution as redefined when the numerical optimization has not converged, and 
 using mean values of the distribution of controller parameters as the final tuned parameters when the numerical optimization has converged. 
 
     
     
       2. The method according to  claim 1  wherein initializing a distribution of controller parameters includes defining a normal distribution with a mean and a standard deviation for each of the parameters. 
     
     
       3. The method according to  claim 1  wherein the controller parameters to be tuned include an input target force vector to the compliance controller model, in six degrees of freedom, for each step in a multi-step assembly path. 
     
     
       4. The method according to  claim 1  wherein the simulations include the compliance controller model which computes robot motions based on a difference between an input target force vector and a feedback contact force vector. 
     
     
       5. The method according to  claim 4  wherein the simulations include contact dynamics between the first and second parts using solid models of the parts, and the feedback contact force vector is computed from the contact dynamics. 
     
     
       6. The method according to  claim 1  wherein the fixed pose deviation of the second part is chosen randomly from a range of pose deviations, and includes a combination of three orthogonal position deviations and three orthogonal orientation deviations. 
     
     
       7. The method according to  claim 1  wherein the cost function value is computed based on a final assembled position error after each simulation, where a lower cost function value designates a smaller final assembled position error. 
     
     
       8. The method according to  claim 1  wherein the numerical optimization has converged when a percentage of the simulations meeting a maximum cost function value criteria exceeds a predefined threshold. 
     
     
       9. The method according to  claim 1  wherein running a numerical optimization includes using Covariance Matrix Adaptation Evolution Strategy (CMA-ES) Optimization, Particle Swarm Optimization, or Bayesian Optimization. 
     
     
       10. The method according to  claim 1  wherein redefining the distribution of controller parameters includes selecting the quantity of the sets of simulations having the lowest average cost function values, and defining a new mean and a new standard deviation for each of the controller parameters based on the controller parameters in the quantity of sets which were selected. 
     
     
       11. The method according to  claim 1  wherein the robotic assembly operation includes one of fitting a non-axisymmetric planar part into a mating aperture, assembling a two-peg part with a two-hole part, or inserting an electrical connector into a mating connector. 
     
     
       12. The method according to  claim 1  further comprising using the final tuned parameters in a robot controller configured with a compliance controller to perform a real-world assembly operation. 
     
     
       13. A computer-implemented method for autonomously tuning controller parameters for a robotic assembly operation, said method comprising using a numerical optimization algorithm to optimize values of the controller parameters, wherein a plurality of random samples are selected from a distribution of the controller parameters, and a set of simulations of the assembly operation is run for each of the samples in the plurality, where each set of simulations includes a plurality of simulations each using a different pose deviation of a fixed part, and where the distribution of the controller parameters is redefined based on a subset of the sets of simulations having best assembly performance until a convergence criteria is met. 
     
     
       14. The method according to  claim 13  wherein the controller parameters to be tuned include an input target force vector to a compliance controller model, in six degrees of freedom, for each step in a multi-step assembly path, where the simulations include the compliance controller model which computes robot motions based on a difference between the input target force vector and a feedback contact force vector, and the simulations include contact dynamics between parts being assembled using solid models of the parts and the feedback contact force vector is computed from the contact dynamics. 
     
     
       15. A system for performing a robotic assembly operation of a first part with a second part, said system comprising:
 a computer with a processor and memory, said computer being configured for autonomously tuning controller parameters for the robotic assembly operation including using a numerical optimization algorithm to optimize values of the controller parameters, wherein a plurality of random samples are selected from a distribution of the controller parameters, and a set of simulations of the assembly operation is run for each of the samples in the plurality, where each set of simulations includes a plurality of simulations each using a different fixed pose deviation of the second part, and where the distribution of the controller parameters is redefined based on a subset of the sets of simulations having best assembly performance until a convergence criteria is met; and 
 a robot controller controlling a robot, said controller being configured with a compliance controller to perform the robotic assembly operation, where the compliance controller uses the optimized values of the controller parameters from the computer. 
 
     
     
       16. The system according to  claim 15  wherein the simulations include a compliance controller model of a robot performing the assembly operation. 
     
     
       17. The system according to  claim 16  wherein the controller parameters to be tuned include an input target force vector to the compliance controller model, in six degrees of freedom, for each step in a multi-step assembly path. 
     
     
       18. The system according to  claim 17  wherein the compliance controller model computes robot motions based on a difference between the input target force vector and a feedback contact force vector. 
     
     
       19. The system according to  claim 18  wherein the simulations include contact dynamics between the first and second parts using solid models of the parts, and the feedback contact force vector is computed from the contact dynamics. 
     
     
       20. The system according to  claim 15  wherein the fixed pose deviation of the second part is chosen randomly from a range of pose deviations, and includes a combination of three orthogonal position deviations and three orthogonal orientation deviations. 
     
     
       21. The system according to  claim 15  wherein the subset of the sets of simulations having best assembly performance is determined by a cost function value computed based on a final assembled position error after each simulation, where a lower cost function value designates a smaller error. 
     
     
       22. The system according to  claim 15  wherein the convergence criteria is met when a percentage of the simulations meeting a maximum cost function value criteria exceeds a predefined threshold. 
     
     
       23. The system according to  claim 15  wherein using a numerical optimization algorithm includes using Covariance Matrix Adaptation Evolution Strategy (CMA-ES) Optimization. 
     
     
       24. The system according to  claim 15  wherein the distribution of the controller parameters is redefined by selecting a quantity of the sets of simulations having lowest average cost function values, and defining a new mean and a new standard deviation for each of the controller parameters based on the controller parameters in the quantity of sets which were selected. 
     
     
       25. The system according to  claim 15  wherein the robotic assembly operation includes one of fitting a non-axisymmetric planar part into a mating aperture, assembling a two-peg part with a two-hole part, or inserting an electrical connector into a mating connector.